Robust quadrotor control with accurate depth estimation and obstacle avoidance is paramount for agile flight in a dynamic environment. This paper presents initial work to establish software and hardware architecture in developing a completely autonomous high-speed racing quadrotor with onboard processing on a 250-size frame capable of indoor flight in a GPS denied environment. The objective of this paper is

1) Develop software and hardware architecture needed for onboard sensing and computation.

2) Implement robust algorithms for localization, mapping, learning, and decision making.